Troubleshooting

Improving Oracle OLTP database performance with Dell Fluid Cache for DAS
17
Performance graph: Total CPU Utilization Figure 7.
Figure 7 shows, that in the case of the Fluid Cache based solution, as the user load keeps increasing,
the total CPU utilization also keeps increasing. In the Fluid Cache based solution, the cache pool
configured on the Dell Express Flash PCIe SSDs is transparent to the database application. As a result,
with caching enabled on the backend store, the database reads, re-reads, and writes are transparently
managed by the Fluid Cache software.
With the Fluid Cache based solution configured in write-back mode, the modified or the dirty data
flushed from the SGA buffer cache in the main-memory by the database is written to the Fluid Cache
for DAS cache pool. Fluid Cache transfers this cached dirty data back to the backend storage at a later
time. Without Fluid Cache, this dirty data would be flushed directly back to the spinning disk. As a
result, from the database point of view, most of the physical read and write transactions are
completed with very low latency even at high user loads in a Fluid Cache based solution. This
attributes to the increased CPU utilization. However, as seen in Figure 7, the CPU eventually hits 100%
utilization that results in response times crossing the two-second threshold. Thus, the CPU eventually
becomes the bottleneck in case of the Fluid Cache based solution.
0
20
40
60
80
100
%
Increasing User Load -->
OLTP (R720) - Total CPU Utilization
Baseline Fluid Cache